Open Journal of Political Science
2013. Vol.3, No.1, 30-38
Published Online January 2013 in SciRes (http://www.scirp.org/journal/ojps) http://dx.doi.org/10.4236/ojps.2013.31005
30
Lazy or Greedy? Impact of Xenophobic Beliefs on Natives’
Attitudes towards Redistribution
Raul Magni Berton
University of Grenoble, Science po Grenoble, PACTE, Grenoble, France
Received October 22nd, 2012; revised November 26th, 2012; accepted December 13th, 2012
Several recent accounts have shown that anti-immigrant feeling among citizens seems to reduce the sup-
port for the welfare state. As a consequence, the rise of immigration could produce a deep change in in-
dustrialized countries’ social security systems. This paper provides evidence that support for redistribu-
tion is not decreased by generic xenophobia, but by a specific kind of xenophobic belief. It also shows
that some other xenophobic beliefs tend rather to produce a demand for governmental protection pro-
grams. Based on a multivariate analysis on individual and contextual French data, findings show that the
support for social protection programs is positively related to the fear of competition from immigrants and
negatively with the fear that immigrants strain the welfare state. This result can be generalized to other
countries where “redistributive xenophobia” is much more widespread.
Keywords: Immigration; Xenophobia; Welfare State; France; Redistribution
Introduction
The issues linked to immigration have been multiplying in
western democracies. During the last decade, security, intoler-
ance, national identity, unemployment or the welfare state, have
all been associated with the rise in immigration. Particularly,
several studies have provided evidence of a negative correlation
between income redistribution and the rise in immigration
(Finseraas, 2008; Roemer, Lee, & Van Der Straeten, 2007;
Senik et al., 2009; Eger, 2010). The general explanation for this
relationship is that the welfare state implies a certain concept of
community: the shared burdens and benefits of the welfare state
are acceptable so long as everyone shares the same features
(McGhee & Neiman, 2010). So, citizens tend to show less soli-
darity with “different” people and wish to decrease the generos-
ity of state benefits. Assuming that diversity depends on lin-
guistic, ethnic or racial features, some studies suggest that frac-
tionalization reduces the size of the welfare state (Mueller &
Murrel, 1986; James, 1993; Alesina & La Ferrara, 2005). This
fact leads Alesina and Glaeser (2004: p. 11) to predict that “if
Europe becomes more heterogeneous due to immigration, eth-
nic divisions will be used to challenge the generous welfare
state”.
However, other recent studies have shown that such a causal
link is not so robust, and other factors have to be considered
(Crepaz, 2008; Johnston et al., 2010). Particularly, reducing
solidarity does not automatically imply a decrease in support
for social protection.
Among the arguments put forward, Alesina and Glaser (2004)
noted that if the “different” minority is on average richer than
other people, probably redistributive policies will be massively
supported. This positive relationship between minorities and
redistribution was observed in the first half of the twentieth
century, in Belgium. Before 1970, the Walloons were both
richer and less numerous than the Flanders. This can explain
the high level of redistribution in this country, contrary to what
happens in other linguistically fragmented countries, such as
Switzerland or Canada. So, the absence of feelings of solidarity
towards immigrants can lead to increasing support for the Wel-
fare State if redistribution allows the majority group to be fa-
voured, or protected from minorities (Luttmer, 2001).
The aim of this paper is to clarify the link between beliefs
about immigrants and preferences for redistribution. I will ar-
gue that there is no causality between general attitudes towards
immigration and support for the welfare state, but some specific
beliefs about immigrants cause specific attitudes towards social
protection. Particularly, beliefs about immigrants can be redis-
tribution-oriented or redistribution-neutral. They are redistribu-
tion-neutral if the belief does not lead individuals to have a
particular opinion about redistribution. On another hand, the
beliefs about immigrants are redistribution-oriented if they
bring people to a particular opinion about redistribution. The re-
distribution-oriented beliefs are psychologically linked to a re-
distributive preference. Moreover, they may be positively or ne-
gatively oriented. According to the Alesina and Glaeser (2004)
study, for example, believing that immigrants are rich is posi-
tively oriented to redistribution, whereas perceiving them as
poor produces a negative opinion of redistribution.
The case studied in this article is France in 2008. France is
one of the top three countries receiving the highest number of
immigrants each year (with Germany and the UK), and its total
taxation as a percentage of GDP is much higher than both other
countries and among the highest in Europe according to the
OECD Revenue Statistics in 2008. Politically, the traditional
and the extreme right have paid considerable attention to this
issue. In 2002 the extreme right-wing candidate for the presi-
dential election came second. In 2004 a law was voted banning
Muslim women from wearing veils—as an ostentatious reli-
gious sign—in public schools and other public buildings. In
2007 the Ministry of National Identity was set up. These events
are the most salient examples to illustrate how important immi-
gration and diversity issues are. Finally, previous studies have

R. MAGNI BERTON
shown that immigration in France reduces voter preferences for
expanding the size of the public sector (Roemer & Van Der
Straeten, 2005), even if public opinion remains strongly at-
tached to the welfare state (Bréchon & Tchernia, 2009).
It is possible to formulate the purpose of this article as a
study of attitudes towards social protection focusing on the
immigration issue. The aim is not to provide a complete model
to explain preferences for redistribution, but to evaluate in
which way immigration can influence these preferences. I will
proceed in four steps. The first section develops the theoretical
approach and the main testable hypotheses. The second section
presents the empirical data and the estimation strategy used to
test both hypotheses. The third section provides the main find-
ings about the link between beliefs about immigrants and pref-
erences for redistribution. Finally, the fourth section provides a
discussion of these results.
Theoretical Account
Redistribution-Oriented Beliefs
The general assumption of this paper is that citizens—or a
proportion of them—show less solidarity with immigrants than
with one another (Habyarimana et al., 2007). I also consider
that the proportion of people showing more solidarity with im-
migrants than with fellow citizens is negligible. Thus, even if
most people are as concerned by immigrants as by fellow citi-
zens, politically, the minority who is the least favourable to-
wards immigrants can decisively influence governmental poli-
cies. This paper will focus the analysis on this minority.
As I pointed out above, beliefs about immigrants could in-
fluence the attitudes towards redistribution in two different
ways. On the one hand, if people believe that immigrants are
unemployed, they will infer that immigrants take advantage of
the welfare system. If citizens don’t wish to pay for immigrants,
they will reduce their support for the welfare system. On the
other hand, if people believe that immigrants are employed—
and perhaps disposed to work a great deal for a low income—
they would then tend to deduce that immigrants will take their
jobs away. In this case, they will feel protected by the welfare
system and will approve of it. Figure 1 graphically summarizes
this double causal relationship.
Column A shows a psychological pattern leading people to
demand more state responsibility in the citizens’ welfare. The
Immigrants are working
Fellow citizens lose their jobs
and need the welfare state
Immigrants are not working
Immigrants have no jobs and
need the welfare state
The state should take more
responsibility to ensure that
everyone is provided for
Individuals should take more
responsibility for providing for
citizens
(a) (b)
Figure 1.
Opposite causalities linking attitudes towards immigrants and prefer-
ences for redistribution.
key beliefs—immigrants are hard working and therefore tend to
take jobs away from local citizens—are called here redistribu-
tive xenophobia. The redistributive impact of immigration is a
particular case of the broader effect of the risk of future income
loss on preferences for redistribution (Cusack et al., 2006; Iver-
sen & Soskice, 2001; Moene & Wallerstein, 2001). Some citi-
zens might believe that immigrants pose a threat to current or
future income and thus become more supportive of government
responsibility.
Column B sets out the redistribution-averse psychological
pattern: if immigrants are perceived as chronically unemployed
—or simply poorer—this can lead to increased taxes for redis-
tribution. As a result, some citizens may demand less state re-
sponsibility in the citizens’ welfare. This mechanism leads to
the expected negative correlation between support for the wel-
fare state and the rise in immigration described above.
Finally, Figure 1 uses the government responsibility as a
proxy for redistribution, social protection and the welfare state.
Although these concepts are slightly different, they all suppose
some state responsibility in citizens’ living conditions. State (vs
individual) responsibility for people’s needs can imply some
forms of unredistributive welfare as well as strictly redistribu-
tive measures. In this article, the concepts of welfare or redis-
tribution refer to the general idea of state responsibility for citi-
zens’ welfare.
Data and Methodology
In order to test both causality patterns, I have used the latest
wave of the European Values Survey conducted in 2008 in
France, partly with a random sampling (n = 1501) and partly
with a stratified sampling (n = 1570). Moreover, with respect to
the available EVS data, some specific questions have been in-
troduced (Bréchon & Tchernia, 2009)1. Since this study only
concerns the natives’ attitudes, I eliminated people who do not
have French nationality, reducing the sample to 2958 respon-
dents.
In addiction to individual data, I used aggregate data pro-
vided by the INSEE (the French Institute for statistics and eco-
nomic data). These contextual data relate to the smallest French
administrative scale i.e. the county (département). Into our
sample, the smallest county has about 200,000 inhabitants and
the largest about 2,500,000, due to the presence of metropolis.
Among the 96 metropolitan French counties, the French re-
spondents of WVS live in 86 different counties. On average, a
county represents 1.16 percent of the respondents (i.e. around
36 respondents). The least represented county makes up 0.26
percent of the survey and the most represented 5.21 percent.
The distribution of the survey across the counties is related to
the demographic weight of the counties in the true French po-
pulation.
Key Variables
The key variables correspond to the second and the third
steps of Figure 1. For the third step—the dependent variable—
the question was formulated in identical terms in the question-
1The French sample of the European Values Survey data has been chosen
for two reasons. First—contrary to the European Social Survey, for exam-
ple—it allows the identification of the county scale in which respondents
live. Second, it offers a large variety of questions on values, especially in
questions linked to immigration.

R. MAGNI BERTON
naire. Respondents have to place themselves on a ten point
scale in which 1 corresponds to the statement “Individuals
should take more responsibility for providing for citizens” and
10 corresponds to “The state should take more responsibility to
ensure that everyone is provided for”. Thus, the higher this
indicator is, the more people support state responsibility for
needs (see Table 1). In this first analysis, this indicator will be
the dependent variable.
The second group of variables—referring to the second step-
measures specific beliefs about immigration and is the main
independent variable. For the first time in the European Values
Surveys, five questions on the link between immigration and
the welfare state were asked. These questions are also proposed
with a ten point scale, and two of them test the model, asking
whether immigrants take jobs away (10) or not (1) and whether
immigrants are a strain on the welfare system (10) or not (1). It
is important to note that the correlation between both variables
is high (r = .55). This could indicate that, contrary to what I had
assumed, many people answer both questions in a similar way.
In fact, there is a set of people that are systematically opposed
or systematically favourable to immigrants. So, they believe
that whatever immigrants do, they are a threat, or a resource,
for the society. For example, the most direct question is whe-
ther immigrants will become a threat to society or not. This
question is highly correlated with the other more specific ques-
tions such as those about jobs (r = .63) and those about the
welfare state (r = .74). Finally, correlating the questions about
immigration (including those about crime and cultural life,
described below), we can observe that the statistical relation-
ship between the two independent variables considered here is
the weakest. Assuming that people who are either positively or
negatively disposed towards immigrants would answer both
questions in the same way, it is interesting to understand why
more than 2/3 of the sample answered the two questions differ-
ently and what the consequences of this fact are at the aggre-
gate level.
This differential answering has also been measured with a
ternary variable in which respondents are coded 0 if they an-
swer both questions in the same way. These respondents are not
sensitive to any particular issue. When individuals are coded
with 1, they agree more with the idea that immigrants strain the
welfare system than that they take jobs away (WS sensitive
hereafter). This group is expected to be redistribution-averse
and it corresponds to column B in Figure 1. Finally, people
coded with 2 are more afraid of losing jobs due to immigration.
They are Job sensitive and they should support social protection.
This measure does not consider the level of fear of immigrants,
Table 1.
Description of the key variables.
Mean Std. Err.
State vs Individual responsibility 6.334 .045
Strain the WF 5.097 .050
Take jobs away 3.799 .051
N %
Job sensitive 474 15.43
No sensitive 907 29.53
WS sensitive 1690 55.03
because people answering 10 for both questions have the same
score as the individuals answering 1 for both questions. More-
over, a respondent who puts 7 for welfare strain and 8 for jobs
is definitely classified as job sensitive—as are respondents
putting 0 on welfare strain and 10 on jobs. Therefore, this indi-
cator captures the preference, but not the intensity of this pref-
erence. Basically, it supposes that perceiving any difference
between both issues is fundamentally different from perceiving
a higher or a lower difference. This indicator is adapted to mea-
sure why people are more sensitive to one issue than another,
beyond their degree of xenophobia or their particular sensitivity
to one issue. Table 1 describes these main variables.
Moreover, the respondents had the possibility to give their
opinion on two other specific questions about the way in which
immigration could threaten them. Both questions were formu-
lated on a ten point scale. Respondents could say if immigrants
undermine cultural life in France (10 = totally agree), called
Sap culture. They could also agree with the idea that immi-
grants increase crime (10 = totally agree). These two measure-
ments of xenophobia are expected to be redistribution-neutral,
i.e. they should not have any psychological impact on prefer-
ences for redistribution.
Finally, I have also taken into account a broader negative at-
titude towards immigrants that is not related to a specific belief.
Respondents can strongly agree (1) or strongly disagree (5)
with the statement: “Today in France, there are too many im-
migrants”. This variable—called Tolerance—measures a sim-
ple attitude concerning the number of immigrants living in
France, but it doesn’t specify the reasons for which there are
too many immigrants or not.
Control Variables
Above all, I have to test that beliefs about immigration is a
cause, and not an effect, of attitudes towards state protection. In
order to identify the specific impact of beliefs about immigra-
tion on state protection, the general ideological predisposition
towards free market politics is controlled. Negative attitudes
towards the free market (hereafter Oppose Free Market) are
measured from four questions. The first question concerns com-
petition. Respondents placed themselves on a ten-point scale in
which 1 meant “competition is good. It stimulates people to
work hard and develop new ideas” and 10 was “Competition is
harmful, it brings out the worst in people”. The second question
is about private property. In the ten-point scale, 1 corresponds
to the opinion: “private ownership of business and industry
should be increased” and 10 to the opinion “government own-
ership of business and industry should be increased”. The last
two questions are specific to the French questionnaire. The first
concerns entrepreneurship. People can agree (1) or not (4) with
the sentence: “entrepreneurship should be increased”. The sec-
ond, based on the same structure, deals with profit. The sen-
tence is “a working economy needs private companies to gen-
erate profits”. The addition of these opinions about competition,
private property, entrepreneurship and profit forms the Oppose
Free Market variable. If beliefs about immigration were an
effect of the attitudes towards the market, their effect should be
captured by this control variable.
I have also controlled for income (annual household income
on 14 categories centred on the median income level), expect-
ing that the richer people are, the less willing they are to pay
taxes, and, inversely, when people are poor, they ask for more
32

R. MAGNI BERTON
redistribution. Both variables measure the impact of selfish be-
liefs, related to the personal economic situation of respondents.
I have taken into account three major demographic charac-
teristics such as Gender (1 = female), Age and Education (in 5
points, adapted to the French system). The two former demo-
graphic variables are expected to have a positive influence:
older people and females should be more likely to support re-
distribution (Blekesaune & Quadagno, 2003; Keely & Tan,
2008).
Finally, two cultural variables are considered: the religious-
ness of the respondent—associated traditionally with a lower
level of support for welfare state, and the presence of parents
born outside France (hereafter Immigrant Parents). This is the
case for 17% of respondents, 9% of which have both parents
born outside France. This variable could influence the relation-
ship between attitudes towards social protection and beliefs
about immigrants, because it can have an impact on both of
them.
Contextual Effects
The relationship between the first and the second step of
Figure 1 can be tested with contextual variables. According to
this, people’s beliefs should be influenced by the fact that im-
migrants are working or not. I assume here that people observe
the reality around them. According to INSEE data, immigrants
are about twice as likely to be unemployed as French citizens in
2008 (15% vs 7%). So, this fact could explain why people tend
to perceive immigrants as straining the Welfare State instead of
taking jobs away. Nevertheless, at the county scale, the ratio of
the unemployment rate of immigrants to that of French citizens
varies from 1.2 (almost the same) to 3.7. So immigrants are
globally more unemployed than French citizens, but it is not
perceived this way in each county. This unemployment ratio is
the contextual variable used here: it corresponds to what natives
perceive around them and it is not sensitive to the absolute level
of unemployment. According to the general hypothesis, in
counties with low-employment levels of migrants according to
natives, people are expected to be afraid to pay for immigrants.
Conversely, in counties with many working immigrants, people
sense higher competition with immigrants in the labour market.
Results
Estimation Strategy
The objective of this section is to test the presumed causal
relationship between beliefs about immigration, described in
the second step in Figure 1, and the tendency to believe that the
state should take more care of citizens, represented by the third
step.
With regard to the dependent variable, Table 2 uses two dif-
ferent estimation strategies. In columns named Oprobit, the
dependent variable is on a ten-point scale. Thus, for these vari-
ables, I have to carry out an ordered probit model with an esti-
mation based on the maximum log-likelihood. The second stra-
tegy, in columns called Probit, is a probit analysis in which the
dependent variable is a dummy. I have considered the responses
from values 1 to 5 as 0, and the values from 6 up to 10, as 1. In
this case the value 1 means “The state should take more re-
sponsibility to ensure that everyone is provided for” and the
value 0 indicates disagreement with this opinion.
Section 3.2 compares the effects of generic attitudes towards
Table 2.
Beliefs about immigrants and support for social protection.
Oprobit 1Oprobit 2 Probit 1 Probit 2Probit 3
Immigrants
Tolerance .046**
(.017) .042* (.022) .034 (.027)
Take jobs away .026**
(.011) .032** (.014)
.034**
(.013)
Strain the WS −.024**
(.011) −.034** (.014)
−.037**
(.013)
Sap culture −.008
(.011) −.002 (.015)
Increase crime −.003
(.011) −.016 (.015)
Other I.D.
Oppose F. market.093**
(.008)
.088**
(.005)
.088**
(.007)
.088**
(.007)
.088**
(.007)
Income −.016**
(.007)
−.015**
(.007)
−.021**
(.009)
−.020**
(.009)
−.021**
(.009)
Gender (female).054
(.041)
.051
(.042)
.051
(.054)
.048
(.055)
.052
(.055)
Age −.002*
(.001)
−.0017
(.0012)
−.0026
(.0017)
.002
(.002)
−.002
(.002)
Education −.053**
(.021)
−.043**
(.021)
−.074**
(.027)
−.066**
(.027)
−.071**
(.028)
Religiousness −.005
(.044)
.013
(.044)
−.030
(.057)
−.013
(.058)
−.021
(.058)
Migrants parents.006
(.034)
.006
(.035)
.001
(.045)
−.003
(.045)
−.003
(.045)
Constant
−1.44**
(.178)
−1.24**
(.196)
−1.40**
(.226)
Number of obs =2531 2519 2531 2519 2516
chi2 = 362** 372** 237** 249** 250**
Pseudo R2 = .03 .03 .07 .08 .08
Note: **p < .05, *p < .1.
immigration, on support for social protection, with those of spe-
cific beliefs. It shows that the analysis based on specific beliefs
is more precise and offers quite different results. Section 3.3
focuses on the impact of unemployment ratio on the support for
social protection and analyzes whether the sensitivity for spe-
cific immigration issues mediates such a relationship.
Beliefs about Immigrants
According to classic literature, negative attitudes towards im-
migration reduce support for the welfare state. Columns Opro-
bit 1 and Probit 1 confirm this prediction: global tolerance to
immigrants positively influences support for social protection,
independently to attitudes towards the free market. Also, the
impact is less significant when the indicator of social protection
is a dummy. Notice that education and income have a negative
impact on support for social protection.
This estimation must be compared with those represented in
columns Oprobit 2 and Probit 2, where the variable Tolerance

R. MAGNI BERTON
is replaced with specific beliefs about immigrants. In these esti-
mations, two specific beliefs have no significant impact on
support for social protection (immigrants sap culture and in-
crease crime), whereas two others have opposing signs. The be-
lief that immigrants strain the welfare state decreases support
for redistribution, whereas believing immigrants steal jobs in-
creases this support.
So, the results concerning specific beliefs infirm the idea that
negative attitudes towards immigration have a negative influ-
ence on support for social protection, since this influence de-
pends on why people fear immigrants. Note that in the Probit 2
model, Take jobs away and Strain the WS coefficients are both
more significant than that of Tolerance. Moreover, when these
three variables are included in the same model (Probit 3), Tol-
erance loses its significance, while Take jobs away and Strain
the WS don’t. Finally, regarding Chi2 and the Pseudo R2, using
specific beliefs slightly increases the relevance of the model.
It must be clarified here why Tolerance has a significant im-
pact, whereas when analyzing specific beliefs such an impact is
equivocal. The reason is that, as Table 1 shows above, there are
in France many more people that think of immigrants as strain-
ing the welfare state instead of stealing jobs. This fact explains
why—regarding attitudes—negative attitudes tend to make
people oppose social protection, but when considering people’s
beliefs, results are less clear.
These results are robust and they hold with different estima-
tions. The collinearity between the four beliefs about immi-
grants does not have an impact on results except in one impor-
tant case: the variable “Take jobs away”, taken without another
“immigrant” variable, maintains its positive influence, but loses
its significance at the .05 level. This variable’s strongest and
most significant impact appears when “Strain the WS” is con-
trolled. To understand this result it is useful to consider the
“attitude” effect. As I said above, 1/3 of the sample answered
both questions in the same way. In France, people viewing im-
migrants as a threat tend to vote for the political parties not
promoting redistributive policies and, inversely, people who are
not afraid of immigration are more supportive of redistributive
parties. Therefore, most people connect redistribution and tol-
erance for immigration in this ideological way (Roemer et al.,
2007). When this effect is captured by another immigration
variable such as Sap Culture or Increase Crime, it is possible to
observe the real effect of the belief that immigrants take jobs
away. This effect clearly encourages greater redistribution. In
particular, comparing the respective coefficients, its impact is
similar to that of the “Strain the WF” variable.
Job and WS Sensitivity and Contextual Effects
The previous section has shown that beliefs, rather than atti-
tudes, matter. But to understand the process that brings about
supporting social protection, Figure 1 has to be empirically
operationalized. According to Figure 1, being more sensitive to
the job issue is caused by the perception that immigrants are
hard working and it produces greater support for social protec-
tion. On the other hand, sensitivity to the welfare state issue is
due to a perception of immigrants as not working and leads to a
decrease in support for social protection. Therefore, indirectly,
the unemployment rate of immigrants can lead to greater or
lesser support for social protection. The main hypothesis here is
that the Unemployment Ratio that people perceive around them
has an influence on the way in which they view the immi-
grants2.
In Table 3, the contextual variable Unemployment Ratio is
added to the model and estimated with clustered standard er-
rors3. Moreover, I have dropped Religiousness and Migrant
Parents, because they were insignificant in the previous model
and they do not affect the results.
Columns Oprobit 1 and Probit 1 describe the indirect effect
of the unemployment ratio on support for social protection. In
both cases, its impact is—as expected—negative: the more
unemployed immigrants there are, compared to natives, the less
people support state responsibility. In the probit model, this
effect is only significant at the 10% level. The other variables
keep the impact observed above.
Let’s now introduce the respondents’ sensitivity. Compared
to job sensitivity, WS sensitivity has a significant negative im-
pact. In the probit model, no sensitivity coefficient is signifi-
cantly different from job and WS sensitivity effects, whereas in
the ordered model there is no significant difference between job
sensitivity and no sensitivity.
More interesting, introducing these variables considerably
weakens the impact of the contextual variable. The Unemploy-
ment Ratio has an impact on support for social protection,
which is mediated by the kind of sensitivity. We can perceive
this through the loss in significance and in the slope described
by the coefficient of the Unemployment Ratio, when the vari-
able related to the type of sensitivity is controlled4.
To sum up, the hypotheses summarized in Figure 1 are con-
firmed. When the immigrant unemployment rate is higher than
that of natives, the native support for social protection decreases.
The more unemployed immigrants there are the less generous
natives are. But this effect becomes considerably weaker when
people’s sensitivities are controlled. In this case, when people
perceive immigrants as straining the welfare state, they wish to
reduce the welfare state. On the contrary, when natives perceive
immigrants as job thieves, they increase their support for social
protection. The insignificance of the contextual variable in this
last case, confirms that its effect is captured by the sensitivity
variables.
Generalizing Results
Overall, the welfare state does not always seem to suffer
from fear of immigrants. Sometimes—when immigrants are
perceived as job thieves, support for the welfare state increases.
Inversely, when people worry about immigrants because they
think they are lazy and take advantage of the welfare state, sup-
port for the welfare state decreases. Other kinds of xenophobic
beliefs—immigrants sap our culture, or produce crime—are
neutral, i.e. not significantly linked with support for redistribu-
tion. In France, immigration has a negative effect on public
spending (Roemer & Van Der Straeten 2005) probably because
currently there are about three times more people who believe
2Of course, this is not a necessary consequence. Indeed, people can wrongly
p
erceive immigrants as working or not working. In this article, only the
impact of the real situation is tested.
3The error terms of the estimation could be correlated with unobserved fea-
tures of the counties. It is possible that the errors are not identically and
independently distributed. To solve this problem, we use a method to correct
the variance of the errors known as “clustering correction”. The corrected
variance related to the counties provides corrected standard error of the
estimated coefficients that do not suffer from heteroscedasticity.
4Note that when the sensitivity variables are controlled, the impact of the
other variables is not affected, except for age. I cannot account for that.
34

R. MAGNI BERTON
Table 3.
Sensitivity for immigration issues, unemployment ratio and support for
social protection.
Oprobit 1 Oprobit 2 Probit 1 Probit 2
Immigrants
Job sensitivity
(excluded)
WS sensitivity −.166**
(.067) −.303** (.080)
No sensitivity −.082
(.076) −.019** (.095)
Contextual
Unemployment ratio −.078**
(.036)
−.069*
(.035)
−.076*
(.046)
−.062
(.045)
Other I.D.
Oppose F. market .089**
(.006)
.089**
(.006)
.089**
(.007)
.089**
(.007)
Income −.011
(.008)
−.011
(.008)
−.016*
(.010)
−.017*
(.010)
Gender (female) .046
(.043)
.045
(.043)
.044
(.053)
.041
(.053)
Age −.002*
(.001)
−.002
(.0012)
−.003*
(.002)
.002
(.002)
Education −.040*
(.022)
−.036*
(.022)
−.062**
(.027)
−.055*
(.028)
Constant
−1.22**
(.204)
−1.07**
(.202)
Number of obs = 2600 2581 2600 2581
chi2 = 233** 245** 212** 271**
Pseudo R2 = .03 .03 .07 .08
Note: **p < .05, *p < .1.
that immigrants are lazy than people who believe they steal jobs
(1690 vs 474 in the survey). But this fact is contingent. An
inversion of these beliefs could lead to the inverse effect on
support for public spending. As the results show, negative
opinions of immigrants can have an effect on support for the
welfare state, but this effect depends on the content of these ne-
gative opinions.
Results have also shown that this kind of redistributive xe-
nophobia is rare in France. But in other parts of the world, the
redistributive xenophobia is more developed. Figure 2 illus-
trates this hypothesis. The World Values Survey of 1999 on 54
countries around the world shows a strong negative correlation
between believing that individuals (instead of the state) should
take more responsibility for providing for themselves (shown
by the mean of answers on the ten point scale described above)
and believing that when jobs are scarce, employers should give
priority to fellow citizens over immigrants (shown in percent-
age). Figure 2 illustrates at the aggregate level the evidence
provided in France at the individual level. The fear of “working
immigrants” increases support for redistribution.
Note that the recent accounts on the relationship between
immigration and redistribution have underlined a negative cau-
sality, confirmed by several empirical tests (Roemer et al., 2007;
Soroka et al., 2006). But such tests focused on countries de-
picted at the bottom-right of Figure 2, especially the US, New
Zealand, Australia and some North European countries. In these
countries, few people have a redistributive xenophobia. Gener-
alizing the findings of this paper, xenophobia decreases prefer-
ences for redistribution in countries where the redistribution-
averse xenophobia has the upper hand. These countries are on
the right of Figure 2. Inversely, xenophobia should increase
preferences for redistribution in more redistributive xenophobic
countries, depicted on the left of Figure 2.
According to the third wave of the World Values Survey,
France is one of the top ten countries asking for less govern-
ment intervention (support for government intervention is low-
est in Sweden) and one of the six bottom countries to believe
that fellow citizens should be employed as a priority. As I have
also shown, the French are much more afraid that immigrants
strain the welfare state than that they take jobs away. All these
data tend to confirm that beliefs about immigrants, in France,
are set in a way that predicts a drop in preference for redistribu-
tion. Of course, redistributive xenophobia also occurs, but con-
siderably less. Similar cases include the US, the UK and several
other western democracies.
Conclusion
The findings have shown two main points: firstly, the causal
link between xenophobia and support for redistribution exists,
but it depends on specific beliefs about immigrants. Second,
these beliefs are partially explained by the effective integration
of immigrants in the labor market. This section analyses the
repercussions of these results on the current situation and on the
literature on attitudes towards immigration.
Let us start with the complex causality linking xenophobia
and support for state responsibility. Some accounts have stated
that a homogeneous community produces strong feelings of
solidarity. Immigration tends to break communities up, thus
weakening solidarity (Habyarimana et al., 2007). This process
would decrease support for redistribution. But, this argument is
challenged by evidence provided here. Feelings of solidarity
towards fellow-citizens can be reinforced if those people are
threatened. So, immigration does not necessarily break com-
munities up, but it can actually reinforce intra-community soli-
darity (Campbell, 2006). The “rally around the flag” (Mueller,
1973) is generally encouraged by politicians to face a foreign
menace, which may be identified as a foreign country, but also
as immigrants. The fear of losing jobs because of immigrants
can be viewed as a foreign threat producing a preference for
redistributing wealth and being protected by government.
As I have also shown, citizens afraid of “lazy immigrants”
tend to decrease their support for redistribution. This result can
also be understood as an effect of intra-group solidarity: people
refuse to let their fellow-citizens pay for immigrants.
Other kinds of xenophobia have no significant impact on re-
distribution. So, the causal link between redistribution and atti-
tudes towards immigration is contingent and depends on spe-
cific beliefs about immigrants. The causal relationship exists,
but only in certain cases. Moreover, when it exists, it can go in
two opposite directions: it can be positive or negative. It is also
possible to reformulate this result, asserting that beliefs about
immigrants have a causal relationship with preferences for re-
distribution, but this is not the caattitudes towards immi- se for

R. MAGNI BERTON
36
France
UK
Italy
Netherlands
Spain
USA
Canada
Japan
Mexi co
S. Africa
Australia
Norw ay
Sweden
Ar g enti na
Fi n land
South Korea
Poland
Swi tze r land
Brazil
Chile
India
Sl o venia
Bulgaria
Romania China
Taiwan
Tu r k ey
Ukrai ne
Russia Peru
Uruguay
Ghana
Mo ld ova
Ge or gi a
Thailand
Indonesia
Viet Nam
Se r b ia
N. Zealand
Egypt
Mo rocc o
Iran
Jordan
Cyprus Guatemala
Hong K.
Trinidad and T.
Malaysia
Burkina F.
Et hi o pi a
Mali
Rwanda
Zambia
Ge r many
0,00%
00
,
00%
When jobs are scarce, employers
should give priority to fellow
citizens over immigrants
(percentage per country)
Individuals (vs. the state) should take more responsibility
for providing for citizens (mean per country)
0.00%
100.00%
3 4 5 6
Figure 2.
Relationship between redistribution-seeking xenophobia (Y) and preference against state intervention (X) in 54 countries.
REFERENCES gration. So, to understand how xenophobia affects redistribu-
tion, it is necessary to explore on which beliefs xenophobia is
based. Alesina, A., & Edward, L. G. (2004). Fighting poverty in the US and in
Europe: A world of difference. New York: Oxford University Press.
doi:10.1093/0199267669.001.0001
These beliefs are presumably affected by the real situation.
When immigrants are working, natives fear for their jobs,
whereas when immigrants are unemployed, natives are afraid of
public spending. According to this point, politicians can indi-
rectly exploit citizens’ fears. Anti-redistributive governments
have an incentive to manage a migratory policy leading to a
high immigrant unemployment rate, in order to make people
perceive immigrants as a strain for the welfare state and reject
redistribution5. On the other hand, pro-redistributive parties
would do better to select immigrants and give them good op-
portunities to work in order to make the local citizens afraid of
losing their jobs and give them a desire to be protected by the
welfare state. This prediction seems quite correct, at least con-
cerning the link between immigration and redistribution: in
most countries, the same political parties defend both immi-
grants and redistribution.
Alesina, A., & Eliana La, F. (2005). Preferences for redistribution in the
land of opportunities. Journal of Public Ec o n om i c s, 89, 897-931.
doi:10.1016/j.jpubeco.2004.05.009
Blekesaune, M., & Quadagno, J. (2003) Public attitudes toward welfare
state policies: A comparative analysis of 24 nations. European So-
ciological Review, 19, 415-427. doi:10.1093/esr/19.5.415
Bréchon, P., & Tchernia, J.-F. (Eds.) (2009). La France à travers ses
valeurs. Paris: Armand Colin.
Campbell, D. E. (2006). Why we vote: How schools and communities
shape our civic life. Princeton, NJ: Princeton University Press.
Crepaz, & Markus M. L. (2008) Trust beyond Borders. Ann Arbor:
University of Michigan Press.
Cusack, T., Iversen, T., & Rehm, P. (2006). Risk at work: The demand
and supply sides of government redistribution. Oxford Review of
Economic Policy, 22, 365-389. doi:10.1093/oxrep/grj022
Eger, M. A. (2010). Even in Sweden: The effect of immigration on
support for welfare state spending. European Sociological Review,
26, 203-217. doi:10.1093/esr/jcp017 Finally, the findings discussed above about the immigration
issue have to be compared with findings provided by the litera-
ture on “racial fractionalization”. Both questions—immigration
and racial homogeneity—are empirically linked, but conceptu-
ally different. In Europe, immigrants are on average racially
different from natives. But, in the US, especially in the twenti-
eth century, there was maybe more racial fractionalization
among natives than between natives and immigrants. In France,
even if immigration is still associated with ethnic diversity, the
percentage of Blacks or Arabs among natives is increasing.
Today, a study on immigration in France can be considered as
tallying with the racial diversity issue. In the future, we could
compare the specific immigration effect and the racial diversity
effect in most European countries.
Finseraas, H. (2008). Immigration and preferences for redistribution:
An empirical analysis of European survey data. Comparative Euro-
pean Politics, 6, 407-431. doi:10.1057/cep.2008.3
James, E. (1993). Why do different countries choose a different pub-
lic-private mix of educational services? Journal of Human Resources,
28, 571-592. doi:10.2307/146161
Johnston, R., Bantinga, K., Kymlicka, W., & Soroka, S. (2010). Na-
tional identity and support for the welfare state. Canadian Journal of
Political Science, 43, 349-377. doi:10.1017/S0008423910000089
Habyarimana, J., Humphreys, M., Posner, D. D., & Weinstein, J. M.
(2007). Why does ethnic diversity undermine public goods provision.
American Political Science Review, 101, 709-725.
doi:10.1017/S0003055407070499
Iversen, T., & Soskice, D. (2001). An asset theory of social policy pref-
erences. American Political Science Review, 95, 875-895.
Keely, L. C., & Chih, M. T. (2008). Understanding preferences for
income redistribution. Journal of Public Economics, 92, 944-961.
doi:10.1016/j.jpubeco.2007.11.006
5Of course, this is a risky strategy. An integration-policy-fiasco could also
undermine the probability of re-election.

R. MAGNI BERTON
Luttmer, E. F. P. (2001). Group loyalty and the taste for redistribution.
Journal of Political Economy, 109, 500-528. doi:10.1086/321019
McGhee, E. M., & Neiman, M. (2010). Concern over immigration and
support for public services. California Journal of Politics and Policy,
2.
Moene, K. O., & Wallerstein, M. (2001). Inequality, social insurance,
and redistribution. American Political Science Review, 95, 859-874.
Mueller, J. E. (1973). War, presidents and public opinion. New York:
Wiley.
Mueller, D. C., & Murrell, P. (1986). Interest groups and the size of
government. Publi c Choice, 48, 125-145. doi:10.1007/BF00179727
Roemer, J. E., & Van der Straeten, K. (2005), Xenophobia and the size
of the public sector in France: A politico-economic analysis. Journal
of Economics, 86, 95-144. doi:10.1007/s00712-005-0142-x
Roemer, J. E., Lee, W., & Van der Straeten, K. (2007). Racism, xeno-
phobia and redistribution, a study of multi-issue politics in advanced
democracies. Cambridge: Harvard University Press.
Senik, C., Stichnoth, H., & Van der Straeten, K. (2009). Immigration
and natives’ attitudes towards the welfare state: Evidence from the
European social survey. Social Indicators Research, 91, 345-370.
doi:10.1007/s11205-008-9342-4
Soroka, S., Banting, K., & Johnston, R. (2006). Immigration and redis-
tribution in a global era. In S. Bowles, P. Bardhan, & M. Wallerstein
(Eds.), Globalization and egalitarian redistribution (pp. 261-288).
Princeton, NJ: Princeton University Press.

R. MAGNI BERTON
Appendix I: Cuts for Ordered Estimations with
Standard Errors (Corrected, in Table 3, with
the Cluster Method, by County)
Cuts for Table 2.
Oprobit 1 Oprobit 2
Cut 1: from 1 to 2 −.17 (.14) −.35 (.15)
Cut 2: from 2 to 3 .18 (.14) −.01 (.15)
Cut 3: from 3 to 4 .72 (.14) .54 (.15)
Cut 4: from 4 to 5 1.07 (.14) .89 (.15)
Cut 5: from 5 to 6 1.57 (.14) 1.40 (.15)
Cut 6: from 6 to 7 1.86 (.14) 1.69 (.15)
Cut 7: from 7 to 8 2.18 (.14) 2.01 (.15)
Cut 8: from 8 to 9 2.71 (.14) 2.53 (.16)
Cut 9: from 9 to 10 2.96 (.15) 2.80 (.16)
Cuts for Table 3.
Oprobit 1 Oprobit 2
Cut 1: from 1 to 2 −.42 (.16) −.49 (.16)
Cut 2: from 2 to 3 −.08 (.16) −.15 (.16)
Cut 3: from 3 to 4 .47 (.16) .40 (.16)
Cut 4: from 4 to 5 .82 (.17) .75 (.17)
Cut 5: from 5 to 6 1.32 (.17) 1.26 (.17)
Cut 6: from 6 to 7 1.61 (.18) 1.55 (.18)
Cut 7: from 7 to 8 1.93 (.18) 1.88 (.18)
Cut 8: from 8 to 9 2.45 (.19) 2.39 (.19)
Cut 9: from 9 to 10 2.72 (.20) 2.66 (.19)
Appendix II: Description of the
Control Variables
Individual VariablesMean Std. Err.
Sap Culture 4.044 .052
Increase Crime 4.187 .049
Tolerance 2.805 .025
Oppose Free Market5.092 .045
Income 6.603 .063
Gender (female) .54 .009
Age 48.61 .342
Education 1.977 .023
Religiousness .422 .009
Migrant Parents .259 .011
Contextual Variable
Unemployment Ratio2113 .009
38